Researcher, Agent Safety, Training and Evaluations
Core
Train and evaluate frontier AI models to reduce harmful or misaligned agent actions, turning real-world failures into repeatable safety signals.
Role type
Senior research engineer (agent safety & evaluation)
Builds
Scalable measurement, data-processing, and evaluation systems for AI agents
Domain
AI safety, frontier model research, agent alignment
Deliverable
production ML models
Required skills
research engineering, ML engineering, quantitative research, applied model research, experimentation, data processing, evaluation, infrastructure
Preferred skills
intuition for modern frontier-model research, ability to own ambiguous projects end to end
Technologies
frontier models, evaluation systems, data-processing pipelines
Responsibilities
Train and evaluate frontier models to reduce harmful actions; mine incidents and build scalable measurement systems; collaborate with post-training and capabilities partners to ship mitigations
Seniority
Senior, hands-on IC